Planetside Software Forums

General => Image Sharing => Topic started by: Matt on November 13, 2009, 02:51:58 PM

Title: Altocumulus R&D
Post by: Matt on November 13, 2009, 02:51:58 PM
Inspired by a nice Vue image that was published recently. This was rendered in the upcoming Terragen 2.1, but a similar approximation is possible with 2.0. Render time 6 minutes 29 seconds on 8 threads, i7 920, 2.6Ghz.

Files now available here: http://forums.planetside.co.uk/index.php?topic=3691.msg85522#msg85522

Matt
Title: Re: Altocumulus R&D
Post by: littlecannon on November 13, 2009, 02:57:51 PM
Very nice Matt. 2.1 eh, I can't wait. Renders do come out nice with the new version don't they.
Cheers,
Simon.
Title: Re: Altocumulus R&D
Post by: Seth on November 13, 2009, 03:23:18 PM
nice altocumulus indeed.
and nice to see you rendering something Matt ! ^^
Title: Re: Altocumulus R&D
Post by: Kadri on November 13, 2009, 03:31:53 PM
Nice render Matt.
Are there any changes for rendering clouds you can share with us at this time Matt ?

Kadri.
Title: Re: Altocumulus R&D
Post by: Zylot on November 13, 2009, 03:44:00 PM
Yeah, looks good.  We need to post more VUE renders so Matt gives us more relevant teasers ;)

I echo the point to:  If you can tell us what is different in the clouds, new features or methods or nodes or noise types or anything.
Title: Re: Altocumulus R&D
Post by: domdib on November 13, 2009, 03:44:42 PM
Looks great.
Title: Re: Altocumulus R&D
Post by: Matt on November 13, 2009, 03:45:17 PM
Kadri, there is one new cloud parameter that allows the samples to skip over large volumes of empty space, but I didn't use it here. With this render, the main difference between how it renders in 2.0 and 2.1 is that I ray-traced the pixels, rather than breaking the background sphere into micropolygons and shading those. This is done using a new option called "ray trace everything". For most landscape scenes you still need to use the old rendering mode, but it wouldn't have made a huge difference on this scene anyway.

So basically these clouds are possible in 2.0. I'll post an example scene soon.

Apart from that, not much has changed in the clouds for 2.1. The biggest thing we've been working on is "ray trace objects", which improves quality and speed of your imported objects and vegetation, especially populations. It also means we avoid some of the bugs in the old population rendering engine which we think may have caused many of the render errors people were seeing.
Title: Re: Altocumulus R&D
Post by: Kadri on November 13, 2009, 03:54:14 PM
Wow thanks Matt.
I have to read it more than ones to really understand .
Matt if i may ask another question , by render errors do you mean the black triangles or errors on imported objects (obj ...) ?

I think your new method is really faster?
I have seen the render engine wasting times  behind object rendering . İs this new method affecting this render behavior too?

Thanks.

Kadri.
Title: Re: Altocumulus R&D
Post by: Matt on November 13, 2009, 04:14:10 PM
Quote from: Kadri on November 13, 2009, 03:54:14 PM
Matt if i may ask another question , by render errors do you mean the black triangles or errors on imported objects (obj ...) ?

We think it may avoid bugs that caused some buckets not to complete when they contained populations. There were some weird issues we couldn't figure out with the old renderer. The black triangles are a different problem which we're investigating right now.

Quote
I think your new method is really faster?
I have sean the render engine wasting times  behind object rendering . İs this new method affecting this render behavior too?

Rendering the terrain using ray tracing has its own problems. Usually terrain is more efficient and renders with higher quality with the old method, despite the overlap. For that reason, the default settings in 2.1 are to use a hybrid approach, using ray tracing for objects and using the old micropolygon rasteriser for the terrain, background and other displacement-friendly surfaces.

Matt
Title: Re: Altocumulus R&D
Post by: Kadri on November 13, 2009, 04:28:09 PM
I think i know what you mean with  " Rendering the terrain using ray tracing has its own problems. Usually terrain is more efficient and renders with higher quality with the old method, despite the overlap. For that reason, the default settings in 2.1 are to use a hybrid approach, using ray tracing for objects and using the old micropolygon rasteriser for the terrain, background and other displacement-friendly surfaces. "  I know the behavior of the renderer on landscapes sometimes(back rendering?) But my english might mislead me.

My question was more like  : Think of a big imported object (obj) in front and very big (1/3 of the render size or so) the render engine computes the back side of the object(the landscape or clouds behind) too.

No more questions tonight i promise   :D

Thanks Matt.

Kadri.
Title: Re: Altocumulus R&D
Post by: CCC on November 13, 2009, 04:28:40 PM
All that is really missing for clouds are two things i think. Multiple internal light scattering and a way to create very finely detailed edge feathering effects.    ;D
Title: Re: Altocumulus R&D
Post by: domdib on November 13, 2009, 04:42:20 PM
Just out of curiosity, why was internal scattering removed completely, rather than retained as an option with the caveat that it increases render time?
Title: Re: Altocumulus R&D
Post by: Matt on November 13, 2009, 04:54:47 PM
Quote from: domdib on November 13, 2009, 04:42:20 PM
Just out of curiosity, why was internal scattering removed completely, rather than retained as an option with the caveat that it increases render time?

It does not increase render time. Fake internal scattering is a local point-based cheat that changes the lighting contribution without expensive calculations. The reason it was removed (only for GI contributions) is because I realised it was wrong in most situations, creating the "oily look" that people were complaining about. It still works with direct lighting (e.g. sunlight), so I didn't want to change the default setting to 0. The easiest thing was to disable its effect on GI contribution until I work out a more realistic implementation. I am trying to make sure that clouds being created now will look similar (but automatically improved) when they are re-rendered in future versions, without having to add too many switches in the GUI to maintain backward/forward compatibility.
Title: Re: Altocumulus R&D
Post by: Gannaingh on November 13, 2009, 05:43:02 PM
Nice clouds! It's good to hear about all the improvements that you guys have been working one. Thank you for all of your hard work; I look forward to the next release  :)
Title: Re: Altocumulus R&D
Post by: Henry Blewer on November 13, 2009, 05:49:41 PM
There is a North Easter south-east of here. The clouds over head here in Binghamton looked very similar; and I was wondering how to make them. I considered using a mask, but now I'll just wait and let you do the work. :-[
Very good render Matt. It is nice to see that you use your software. 8) I bet everyone would like for you to do more of this. (Get an old Pentium 4 HT. It leaves lots of time to program ;D)
Title: Re: Altocumulus R&D
Post by: JimB on November 13, 2009, 07:23:32 PM
That's very tasty.
Title: Re: Altocumulus R&D
Post by: Matt on November 13, 2009, 08:45:39 PM
Thanks :)

njeneb, yeah :) slow computers are OK for final renders while working on something else, but I have no shortage of slow computers...  It's the iterative process of getting to the final render, that's where we all need TG to be faster.
Title: Re: Altocumulus R&D
Post by: Matt on November 13, 2009, 10:10:31 PM
Files now available here: http://forums.planetside.co.uk/index.php?topic=3691.msg85522#msg85522
Title: Re: Altocumulus R&D
Post by: dandelO on November 13, 2009, 11:26:17 PM
Are the cloud quality samples as low in your large render as they are in the file you uploaded, Matt? Really only 0.25 quality but with no acceleration cache? I rendered it out of the box and it looks very clean, no noise at all. And the rays appear to be multi-coloured too, not just lighter coloured shafts that I usually end up with, these resemble the spectrum. This is great, I love that network for the cloud shapes too, I could never have thought of anything like that. :( And it looks so simple.
Now we can all see TG used absolutely correctly(;)), brilliant, cheers!
Title: Re: Altocumulus R&D
Post by: Matt on November 13, 2009, 11:47:34 PM
The cloud samples and atmosphere samples are the same as when I rendered it, but mine was rendered using different render settings in 2.1 so it's difficult to compare directly. I tested it in 2.0 and it seemed reasonably noise free and rendered in a similar time at the same resolution.

Yeah, I think it was cloud quality 0.25. I disabled the acceleration cache because I noticed some problems with it.

Matt
Title: Re: Altocumulus R&D
Post by: MacGyver on November 14, 2009, 05:33:55 AM
Quote from: Seth on November 13, 2009, 03:23:18 PM
nice altocumulus indeed.
and nice to see you rendering something Matt ! ^^

I second that, beautiful clouds :)
Title: Re: Altocumulus R&D
Post by: j meyer on November 14, 2009, 10:08:59 AM
Yep,nice clouds! And thanks for the files,always good to have some
examples done by the man himself ;).
Title: Re: Altocumulus R&D
Post by: inkydigit on November 14, 2009, 10:24:49 AM
thanks Matt!
Title: Re: Altocumulus R&D
Post by: buzzzzz1 on November 14, 2009, 10:47:13 AM
Thank You Matt!  Just inserted these into my Mt. Rainier scene. I'll post the results if it works out.
Title: Re: Altocumulus R&D
Post by: red_planet on November 14, 2009, 11:37:37 AM
looks good..

hope to test on my rig soon.

thanks Matt

Rgds

Chris
Title: Re: Altocumulus R&D
Post by: mogn on November 15, 2009, 12:27:20 AM
OT, if you want to examine e.g. the output from the Colour adjust shader in Matts cloud layer:
Creat a displacement shader, connect the output from the from the adjust shader to input 2 of the
displacement shader. Double click the displacement shader. Right click and select "Open in new window"
If you place the mouse in the new view, you can directly reads the numerical output from the adjust shader
as y value in the view.
Title: Re: Altocumulus R&D
Post by: domdib on November 15, 2009, 11:08:03 AM
With regard to low samples, I've noticed this in a couple of the NWDA packs - it seems to be something you can get away with with thin cloud layers.
Title: Re: Altocumulus R&D
Post by: Oshyan on November 15, 2009, 09:30:05 PM
Quote from: domdib on November 15, 2009, 11:08:03 AM
With regard to low samples, I've noticed this in a couple of the NWDA packs - it seems to be something you can get away with with thin cloud layers.

Correct, the thinner the cloud layer, generally the lower the samples needed for good quality.

- Oshyan
Title: Re: Altocumulus R&D
Post by: choronr on November 15, 2009, 11:23:31 PM
Matt, thank you for the files; but, how do you reseed the cloud pattern?

Bob
Title: Re: Altocumulus R&D
Post by: Matt on November 17, 2009, 06:15:50 PM
Create a Constant Scalar function and connect it to the "seed" input of the perlin 3d function. You can then enter any whole-number value in the Constant Scalar function.

Matt
Title: Re: Altocumulus R&D
Post by: choronr on November 17, 2009, 06:29:19 PM
Thanks Matt, these are the most realistic, 'granular' cloud formations I've seen. We see these both as cirrus (high altitude) and altocumulus in real life ...thanks again!
Title: Re: Altocumulus R&D
Post by: domdib on November 18, 2009, 04:31:40 PM
Being an inveterate tweaker, I played around with these a little, and here's what I came up with.

First, after changing the cloud colour and sun angle + height, a simple adjustment of the colour shader.
Title: Re: Altocumulus R&D
Post by: domdib on November 18, 2009, 04:33:08 PM
Second, changed the noise flavour to ridges and added 0.5 warp.
Title: Re: Altocumulus R&D
Post by: domdib on November 18, 2009, 04:35:03 PM
Finally, noise is Perlin Mix 2 and Noise variation 2.
Title: Re: Altocumulus R&D
Post by: FrankB on November 18, 2009, 04:49:03 PM
nice :)
Title: Re: Altocumulus R&D
Post by: Matt on November 18, 2009, 06:24:19 PM
Good variations.
Title: Re: Altocumulus R&D
Post by: Henry Blewer on November 19, 2009, 08:25:05 AM
They don't quite look natural to me, but who cares. They are absolutely stunning! Great job.
Title: Re: Altocumulus R&D
Post by: domdib on November 19, 2009, 04:28:51 PM
Thanks, guys!  A few more... (that's both the joy and the problem with clouds - so many things you can tweak!). You can work out how I did them by decoding the file name  ;)
Title: Re: Altocumulus R&D
Post by: domdib on November 19, 2009, 04:29:44 PM
No.2
Title: Re: Altocumulus R&D
Post by: domdib on November 19, 2009, 04:30:33 PM
And the last one.
Title: Re: Altocumulus R&D
Post by: Henry Blewer on November 19, 2009, 06:30:06 PM
The last one, could be a storm over the horizon.
Title: Re: Altocumulus R&D
Post by: rcallicotte on November 28, 2009, 08:28:14 AM
Wow!  Awesome - great hope for the next version (wonderful!) and your sharing.  Thanks!
Title: Re: Altocumulus R&D
Post by: rcallicotte on October 18, 2010, 07:07:42 PM
Thank you, Matt, for the clips.  This is excellent.